Dominik Kundel reposted the experimental results of GPT-6 Astra directly rewriting model weights
Dominik Kundel reposted Rikiya Takehi's experimental results and said that Astra keeps bringing surprises. The quoted content states that the team had GPT-6 Astra directly modify the weights of Qwen3.5-0.8B to improve its math ability; by changing only 0.016% of the weights, it fixed the 0.8B model's reasoning loops, improving math ability by up to 6x and instruction following by 3x, all without using gradients or training. The authors call this method Weight-Space Programming.
Topics
View original postWe asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: https://t.co/bOWCcFhCVq
View original
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: https://t.co/bOWCcFhCVq
Body
Astra continues to surprise me 🤯